CURVATURE AND CHAOS IN GENERAL-RELATIVITY

We clarify some points about the systems considered by Sota et al. Con trary to the authors' claim for a non-homoclinic kind of chaos, we sho w the chaotic cases they considered are homoclinic in origin. The powe r of local criteria to predict chaos is once more questioned. We find that their local, curvature-based criterion is neither necessary nor s ufficient for the occurrence of chaos. In fact, we argue that a merit of their search for local criteria applied to general relativity is ju st to stress the weakness of locality itself, free of any pathologies related to the motion in effective Riemannian geometries.